toilscript 0.1.4 → 0.1.5

package/std/assembly/bignum/globals.ts

@@ -0,0 +1,546 @@
+import { u128, u256 } from "./integer";
+
+@lazy export var __divmod_quot_hi: u64 = 0;
+@lazy export var __divmod_rem_lo: u64 = 0;
+@lazy export var __divmod_rem_hi: u64 = 0;
+
+// used for returning low and high part of __mulq64, __multi3 etc
+@lazy export var __res128_hi: u64 = 0;
+
+// used for returning 0 or 1
+@lazy export var __carry: u64 = 0;
+@lazy export var __u256carry: u64 = 0;
+@lazy export var __u256carrySub: u64 = 0;
+
+/**
+ * Convert 128-bit unsigned integer to 64-bit float
+ * @param  lo lower  64-bit part of unsigned 128-bit integer
+ * @param  hi higher 64-bit part of unsigned 128-bit integer
+ * @return    64-bit float result
+ */
+// @ts-ignore: decorator
+@global
+export function __floatuntidf(lo: u64, hi: u64): f64 {
+  // __floatuntidf ported from LLVM sources
+  if (!(lo | hi)) return 0.0;
+
+  var v = new u128(lo, hi);
+  var sd = 128 - __clz128(lo, hi);
+  var e = sd - 1;
+
+  if (sd > 53) {
+    if (sd != 55) {
+      if (sd == 54) {
+        v = u128.shl(v, 1);
+      } else {
+        v = u128.or(
+          u128.shr(v, sd - 55),
+          u128.fromBool(u128.and(v, u128.shr(u128.Max, 128 + 55 - sd)).toBool())
+        );
+      }
+    }
+
+    v.lo |= (v.lo & 4) >> 2;
+    v.preInc();
+
+    v = u128.shr(v, 2);
+
+    if (v.lo & (1 << 53)) {
+      v = u128.shr(v, 1);
+      ++e;
+    }
+
+  } else {
+    v = u128.shl(v, 53 - sd);
+  }
+
+  var w: u64 = u128.shr(v, 32).lo & 0x000FFFFF;
+  var u: u64 = <u64>(((e + 1023) << 20) | w) << 32;
+  return reinterpret<f64>(u | (v.lo & 0xFFFFFFFF));
+}
+
+
+/**
+ * Adds two 64-bit unsigned integers `a` and `b` along with a carry-in value.
+ * Sets the global `__carry2` variable to indicate whether an overflow occurred.
+ * @param {u64} a - The first 64-bit unsigned integer.
+ * @param {u64} b - The second 64-bit unsigned integer.
+ * @param {u64} carryIn - Carry-in value (0 or 1).
+ * @returns {u64} - The 64-bit result of the addition.
+ */
+@inline
+export function add64Local(a: u64, b: u64, carryIn: u64): u64 {
+  let tmp = a + carryIn;
+  let carry = tmp < a ? 1 : 0;
+  let sum = tmp + b;
+  carry += sum < tmp ? 1 : 0;
+  __u256carry = carry;
+  return sum;
+}
+
+/**
+ * Subtracts the 64-bit unsigned integer `b` from `a` along with a borrow-in value.
+ * Sets the global `__carry2` variable to indicate whether a borrow occurred.
+ * @param {u64} a - The minuend (64-bit unsigned integer).
+ * @param {u64} b - The subtrahend (64-bit unsigned integer).
+ * @param {u64} borrowIn - Borrow-in value (0 or 1).
+ * @returns {u64} - The 64-bit result of the subtraction.
+ */
+@inline
+export function sub64(a: u64, b: u64, borrowIn: u64): u64 {
+  let tmp = a - b;
+  let borrow = tmp > a ? 1 : 0;
+  let diff = tmp - borrowIn;
+  borrow += diff > tmp ? 1 : 0;
+  __u256carrySub = borrow;
+  return diff;
+}
+
+// @ts-ignore: decorator
+@global
+export function __umulh64(a: u64, b: u64): u64 {
+  var u = a & 0xFFFFFFFF;
+  a >>= 32;
+  var v = b & 0xFFFFFFFF;
+  b >>= 32;
+
+  var uv = u * v;
+  uv = a * v + (uv >> 32);
+  var w0 = (u * b) + (uv & 0xFFFFFFFF);
+  return a * b + (uv >> 32) + (w0 >> 32);
+}
+
+// @ts-ignore: decorator
+@global
+export function __umulq64(a: u64, b: u64): u64 {
+  var u = a & 0xFFFFFFFF;
+  a >>= 32;
+  var v = b & 0xFFFFFFFF;
+  b >>= 32;
+
+  var uv = u * v;
+  var w0 = uv & 0xFFFFFFFF;
+  uv = a * v + (uv >>> 32);
+  var w1 = uv >>> 32;
+  uv = u * b + (uv & 0xFFFFFFFF);
+
+  __res128_hi = a * b + w1 + (uv >>> 32);
+  return (uv << 32) | w0;
+}
+
+// __umul64Hop computes (hi * 2^64 + lo) = z + (x * y)
+// @ts-ignore: decorator
+@inline
+export function __umul64Hop(z: u64, x: u64, y: u64): u64 {
+  var lo = __umulq64(x, y);
+  lo = __uadd64(lo, z);
+  var hi = __res128_hi + __carry;
+  __res128_hi = hi;
+  return lo
+}
+
+// __umul64Step computes (hi * 2^64 + lo) = z + (x * y) + carry.
+// @ts-ignore: decorator
+@inline
+export function __umul64Step(z: u64, x: u64, y: u64, carry: u64): u64 {
+  var lo = __umulq64(x, y)
+  lo = __uadd64(lo, carry);
+  var hi = __uadd64(__res128_hi, 0, __carry);
+  lo = __uadd64(lo, z);
+  hi += __carry;
+  __res128_hi = hi;
+  return lo
+}
+
+// __uadd64 returns the sum with carry of x, y and carry: sum = x + y + carry.
+// The carry input must be 0 or 1; otherwise the behavior is undefined.
+// The carryOut output is guaranteed to be 0 or 1.
+// @ts-ignore: decorator
+@inline
+export function __uadd64(x: u64, y: u64, carry: u64 = 0): u64 {
+  var sum = x + y + carry
+  // // The sum will overflow if both top bits are set (x & y) or if one of them
+  // // is (x | y), and a carry from the lower place happened. If such a carry
+  // // happens, the top bit will be 1 + 0 + 1 = 0 (& ~sum).
+  __carry = ((x & y) | ((x | y) & ~sum)) >>> 63
+  return sum;
+}
+
+// 64x64 => 128 bits
+@inline
+function mul64To128(x: u64, y: u64): u128 {
+  let lo = __umulq64(x, y);  // sets __res128_hi
+  let hi = __res128_hi;
+  return new u128(lo, hi);
+}
+
+/**
+ * A correct 256×256 -> 256 multiply (mod 2^256).
+ * We do standard "schoolbook" multiplication for each 64-bit limb,
+ * skipping products that shift above 255 bits (they vanish mod 2^256).
+ */
+@global
+export function __mul256(
+  ax0: u64, ax1: u64, ax2: u64, ax3: u64,
+  bx0: u64, bx1: u64, bx2: u64, bx3: u64
+): u256 {
+  let result = new u256();
+  let a = [ax0, ax1, ax2, ax3];
+  let b = [bx0, bx1, bx2, bx3];
+
+  // For each (i,j), partial = a[i]*b[j], shift = 64*(i+j).
+  // If shift >= 256, that partial is 0 mod 2^256.
+  // Otherwise shift partial, add to result.
+  for (let i = 0; i < 4; i++) {
+    for (let j = 0; j < 4; j++) {
+      let shift = (i + j) << 6; // 64*(i+j)
+      if (shift >= 256) continue;
+      let p128 = mul64To128(a[i], b[j]); // 128-bit partial product
+      // convert to u256
+      let part = new u256(p128.lo, p128.hi, 0, 0);
+      // shift left by 'shift' bits
+      if (shift != 0) {
+        part = u256.shl(part, shift);
+      }
+      // add to accumulator
+      result = result + part;
+    }
+  }
+
+  return result;
+}
+
+// @ts-ignore: decorator
+@global
+export function __multi3(al: u64, ah: u64, bl: u64, bh: u64): u64 {
+  var u = al, v = bl;
+  var w: u64, k: u64;
+
+  var u1 = u & 0xFFFFFFFF;
+  u >>= 32;
+  var v1 = v & 0xFFFFFFFF;
+  v >>= 32;
+  var t = u1 * v1;
+  var w1 = t & 0xFFFFFFFF;
+
+  t = u * v1 + (t >> 32);
+  k = t & 0xFFFFFFFF;
+  w = t >> 32;
+  t = u1 * v + k;
+
+  var lo = (t << 32) | w1;
+  var hi = u * v + w;
+  hi += ah * bl;
+  hi += al * bh;
+  hi += t >> 32;
+
+  __res128_hi = hi;
+  return lo;
+}
+
+// @ts-ignore: decorator
+@global
+export function __floatuntfdi(value: f64): u64 {
+  var u = reinterpret<u64>(value);
+
+  // if (value < -1.7014118346046e38) { // -(2^127-1)
+  if (value < reinterpret<f64>(0xC7F0000000000000)) { // -(2^128-1)
+    // __float_u128_hi = <u64>-1; // for i128
+    __res128_hi = 0;
+    // overflow negative
+    return 0;
+    // } else if (value < -9.2233720368547e18) { // -2^63-1 // for i128
+  } else if (value < reinterpret<f64>(0xC3F0000000000000)) { // // -(2^64-1)
+    let lo: u64, hi: u64, m: u64;
+
+    m = (u & 0x000FFFFFFFFFFFFF) | (1 << 52);
+    u = (u & 0x7FFFFFFFFFFFFFFF) >> 52;
+
+    u -= 1075;
+    if (u > 64) {
+      lo = 0;
+      hi = m << (u - 64);
+    } else {
+      lo = m << u;
+      hi = m >> (64 - u);
+    }
+    // Convert to 2's complement for correct negative representation
+    lo = ~lo;
+    hi = ~hi;
+    lo += 1;
+    if (lo == 0) hi += 1; // carry to high part
+
+    __res128_hi = hi;
+    return lo;
+    // } else if (value < 9.2233720368547e18) { // 2^63-1 // for i128
+  } else if (value < reinterpret<f64>(0x43F0000000000000)) { // 2^64-1
+    // __float_u128_hi = (value < 0) ? -1 : 0; // for int
+    __res128_hi = 0;
+    // fit in a u64
+    return <u64>value;
+    // } else if (value < 1.7014118346046e38) {
+  } else if (value < reinterpret<f64>(0x47F0000000000000)) { // 2^128-1
+    let lo: u64, hi: u64, m: u64;
+
+    m = (u & 0x000FFFFFFFFFFFFF) | (1 << 52);
+    u = (u & 0x7FFFFFFFFFFFFFFF) >> 52;
+    u -= 1075;
+    if (u > 64) {
+      lo = 0;
+      hi = m << (u - 64);
+    } else {
+      lo = m << u;
+      hi = m >> (64 - u);
+    }
+    __res128_hi = hi;
+    return lo;
+  } else {
+    // overflow positive
+    __res128_hi = <u64>-1; // 0x7FFFFFFFFFFFFFFF for i128
+    return <u64>-1;
+  }
+}
+
+/**
+ * Count leading zeros in a 64-bit unsigned integer `x`, returning i32 in [0..64].
+ * If x == 0, returns 64.
+ */
+function clz64(x: u64): i32 {
+  if (x == 0) return 64;
+
+  let n: i32 = 0;
+  // Check high half [ bits 63..32 ]
+  if ((x & 0xFFFFFFFF00000000) == 0) {
+    n += 32;
+    x <<= 32; // shift left so next checks are for the upper bits
+  }
+  // Check bits [63..48]
+  if ((x & 0xFFFF000000000000) == 0) {
+    n += 16;
+    x <<= 16;
+  }
+  // Check bits [63..56]
+  if ((x & 0xFF00000000000000) == 0) {
+    n += 8;
+    x <<= 8;
+  }
+  // Check bits [63..60]
+  if ((x & 0xF000000000000000) == 0) {
+    n += 4;
+    x <<= 4;
+  }
+  // Check bits [63..62]
+  if ((x & 0xC000000000000000) == 0) {
+    n += 2;
+    x <<= 2;
+  }
+  // Check bit [63]
+  if ((x & 0x8000000000000000) == 0) {
+    n += 1;
+  }
+  return n;
+}
+
+/**
+ * Count trailing zeros in a 64-bit unsigned integer `x`, returning i32 in [0..64].
+ * If x == 0, returns 64.
+ */
+function ctz64(x: u64): i32 {
+  if (x == 0) return 64;
+
+  let n: i32 = 0;
+  // Check low half [bits 31..0]
+  if ((x & 0xFFFFFFFF) == 0) {
+    n += 32;
+    x >>= 32;
+  }
+  // Check bits [15..0]
+  if ((x & 0xFFFF) == 0) {
+    n += 16;
+    x >>= 16;
+  }
+  // Check bits [7..0]
+  if ((x & 0xFF) == 0) {
+    n += 8;
+    x >>= 8;
+  }
+  // Check bits [3..0]
+  if ((x & 0xF) == 0) {
+    n += 4;
+    x >>= 4;
+  }
+  // Check bits [1..0]
+  if ((x & 0x3) == 0) {
+    n += 2;
+    x >>= 2;
+  }
+  // Check bit [0]
+  if ((x & 0x1) == 0) {
+    n += 1;
+  }
+  return n;
+}
+
+/**
+ * Count leading zeros in a 128-bit integer [hi:lo], returning i32 in [0..128].
+ * If both hi and lo are 0, returns 128.
+ *
+ * hi is signed in i128, but we interpret it as unsigned here.
+ */
+// @ts-ignore: decorator
+@global @inline
+export function __clz128(lo: u64, hi: i64): i32 {
+  let h: u64 = <u64>hi;  // reinterpret hi as unsigned
+  if (h == 0) {
+    // If hi is 0, the leading zeros are "64 plus however many are in lo"
+    return 64 + <i32>i64.clz(lo);
+  } else {
+    // The top 64 bits are set => just measure their leading zeros
+    return <i32>i64.clz(h);
+  }
+}
+
+/**
+ * Count trailing zeros in a 128-bit integer [hi:lo], returning i32 in [0..128].
+ * If both hi and lo are 0, returns 128.
+ *
+ * For i128 we typically treat hi as signed, but ctz is purely bitwise, so we
+ * can pass it as u64 as well.
+ */
+// @ts-ignore: decorator
+@global @inline
+export function __ctz128(lo: u64, hi: u64): i32 {
+  if (lo == 0) {
+    // Otherwise, ctz is 64 plus ctz(hi)
+    return 64 + <i32>i64.ctz(hi);
+  } else {
+    // If the lower 64 bits are non-zero, measure ctz(lo)
+    return <i32>i64.ctz(lo);
+  }
+}
+
+// @ts-ignore: decorator
+@global
+export function __udivmod128(alo: u64, ahi: u64, blo: u64, bhi: u64): u64 {
+  var bzn = __clz128(blo, bhi); // N
+
+  // b == 0
+  if (bzn == 128) {
+    throw new RangeError("Division by zero"); // division by zero
+  }
+
+  // var azn = __clz128(alo, ahi); // M
+  var btz = __ctz128(blo, bhi); // N
+
+  // a == 0
+  if (!(alo | ahi)) {
+    __divmod_quot_hi = 0;
+    __divmod_rem_lo = 0;
+    __divmod_rem_hi = 0;
+    return 0;
+  }
+
+  // a / 1
+  if (bzn == 127) {
+    __divmod_quot_hi = ahi;
+    __divmod_rem_lo = 0;
+    __divmod_rem_hi = 0;
+    return alo;
+  }
+
+  // a == b
+  if (alo == blo && ahi == bhi) {
+    __divmod_quot_hi = 0;
+    __divmod_rem_lo = 0;
+    __divmod_rem_hi = 0;
+    return 1;
+  }
+
+  if (!(ahi | bhi)) {
+    __divmod_quot_hi = 0;
+    __divmod_rem_hi = 0;
+    // if `b.lo` is power of two
+    if (!(blo & (blo - 1))) {
+      __divmod_rem_lo = alo & (blo - 1);
+      return alo >> btz;
+    } else {
+      let dlo = alo / blo;
+      __divmod_rem_lo = alo - dlo * blo;
+      return dlo;
+    }
+  }
+  return __udivmod128core(alo, ahi, blo, bhi);
+}
+
+function __udivmod128core(alo: u64, ahi: u64, blo: u64, bhi: u64): u64 {
+  var r = new u128(alo, ahi);
+  var d = new u128(blo, bhi);
+  var q = new u128(0, 0);
+
+  // Standard division by shifting
+  var leadingZeros = __clz128(blo, bhi) - __clz128(alo, ahi);
+
+  if (leadingZeros < 0) {
+    __divmod_quot_hi = 0;
+    __divmod_rem_lo = alo;
+    __divmod_rem_hi = ahi;
+    return 0;
+  }
+
+  var dShift = u128.shl(d, leadingZeros);
+
+  for (let i = 0; i <= leadingZeros; i++) {
+    q = u128.shl(q, 1);
+    if (r >= dShift) {
+      r -= dShift;
+      q = u128.add(q, u128.One);
+    }
+    dShift = u128.shr(dShift, 1);
+  }
+
+  __divmod_quot_hi = q.hi;
+  __divmod_rem_lo = r.lo;
+  __divmod_rem_hi = r.hi;
+  return q.lo;
+}
+
+// @ts-ignore: decorator
+@global
+export function __udivmod128_10(lo: u64, hi: u64): u64 {
+  if (hi == 0) {
+    __divmod_quot_hi = 0;
+    let q = lo / 10;
+    __divmod_rem_lo = lo - (q * 10);
+    __divmod_rem_hi = 0;
+    return q;
+  }
+
+  // High part division
+  let q_hi = hi / 10;
+  let r_hi = hi - (q_hi * 10);
+
+  // Low part chunking to avoid 128-bit math
+  // r_hi becomes the high bits for the next division
+  let lo_hi = lo >>> 32;
+  let lo_lo = lo & 0xFFFFFFFF;
+
+  // Divide upper half of lo (plus carry from hi)
+  let t1 = (r_hi << 32) | lo_hi;
+  let q1 = t1 / 10;
+  let r1 = t1 - (q1 * 10);
+
+  // Divide lower half of lo (plus carry from above)
+  let t2 = (r1 << 32) | lo_lo;
+  let q2 = t2 / 10;
+  let r2 = t2 - (q2 * 10);
+
+  // Reassemble quotient
+  // q1 is known to fit in 32 bits because max t1 < 10 * 2^32
+  let q_lo = (q1 << 32) | q2;
+
+  __divmod_quot_hi = q_hi;
+  __divmod_rem_lo = r2;
+  __divmod_rem_hi = 0;
+
+  return q_lo;
+}

package/std/assembly/bignum/index.ts

@@ -0,0 +1 @@
+export * from "./integer";